Tracking without cookies: server-side tracking and the check nobody does

A broken tracking setup never reports itself. No warning appears and no email arrives. The graph keeps going, the numbers look plausible and nobody notices anything. That lasts until someone compares them with the accounts.

That is exactly what has happened at many Dutch online shops and B2B sites over the past few years.

Where it goes wrong

Measuring in the browser relies on three things, and all three have been eroded: the third-party cookie, a script that is allowed to load, and a visitor who gives consent.

• Ad blockers and tracking protection block the tracking script before it can send anything.
• Safari and Firefox cut the lifetime of a cookie down to a matter of days.
• An honest cookie banner means some visitors decline. They disappear from your figures as if they were never there.

So you are still measuring something. You are measuring a sample and treating it as the whole. What drops out is not random either: technical people, Apple users and privacy-conscious people drop out more often than others. If those happen to be your best customers, you are directing your ads away from them.

Why it is worse than a gap in your reporting

If the tracking only fed your reporting, this would be annoying but manageable. You would take the figures with a pinch of salt and carry on.

But the same tracking also feeds the bidding algorithm. Every conversion you don't measure is a lesson your campaign never gets. So you pay twice: once because you don't know, and again because the system doesn't know either and learns the wrong lessons.

What does work

The answer is server-side tracking: you measure on the server instead of in the browser. The idea is simple. Instead of the visitor's browser talking to Google or Meta, your own server does it on their behalf. That server sees the order even when the script is blocked, knows the real amount and can still report a conversion days later.

This does not get around consent, and it shouldn't. If someone doesn't give consent, you don't measure them. What it does solve is the noise: you now actually capture the people you are allowed to measure.

There is work involved. You need somewhere for that server to run, an agreement on which events count and a way to check that the figures are right. Almost everyone skips that last part, and it is the most important.

The check nobody does

Take any week. Count the orders in your own system. Count the conversions in your ad account. Put them side by side.

If the difference is more than a few per cent, you know enough. Repeat the check every month. A setup that is right today will quietly drift again after an update to your online shop or a change to your cookie banner.

A figure you have never checked is really a hunch with a decimal point.

We do this check as standard before any budget goes in. It is not an exciting job, but everything after it depends on it. Building a campaign on tracking you have not checked is like building a house without checking whether the foundations are there.